In recent years, cardiac troponins have attracted great interest as a marker for myocardial injury. However, there are limited data on strategies for use of creatine kinase (CK)-MB and troponin I (cTnI) in clinical practice. We sought to develop a testing strategy using prospectively collected clinical data including serial CK-MB and cTnI levels from 1,051 patients aged > or = 30 years admitted to a teaching hospital for acute chest pain. Diagnostic performance was evaluated for peak values of CK-MB and cTnI obtained during the first 24 hours for the combined end point of acute myocardial infarction and/or major cardiac events within 72 hours. The overall diagnostic accuracy was similar for both cardiac markers alone, and for the combination of cTnI and CK-MB (receiver-operating characteristic curve 0.84, 0.86, and 0.87, respectively). In the multivariate analysis, models including cardiac markers showed that both CK-MB and cTnI added information to clinical data to predict the combined end point, but cTnI added significantly less. Using recursive partitioning analysis, we developed a strategy that would restrict routine cTnI use to patients with normal CK-MB results and findings on the electrocardiogram consistent with ischemia. This strategy would divide patients with suspected myocardial ischemia into 4 groups with risks for the combined end point of 4%, 13%, 26%, and 85%. Thus, cTnI adds information to CK-MB mass and clinical data for predicting major cardiac events, but this contribution is mainly in patients with evidence of myocardial ischemia on their electrocardiograms.